Pdf/Abstracts/Am2010/Poster Presentations/Posterpresentation Satpm

Corrections

BIOPHYSICS AND COMPUTATIONAL BIOLOGY Correction for “Assessing the utility of coevolution-based residue– The authors note that Fig. 1 C and E and the corresponding residue contact predictions in a sequence- and structure-rich era,” legend appeared incorrectly. The corrected ﬁgure and its legend by Hetunandan Kamisetty, Sergey Ovchinnikov, and David Baker, appear below. which appeared in issue 39, September 24, 2013, of Proc Natl Acad Sci USA (110:15674–15679; ﬁrst published September 5, 2013; 10.1073/pnas.1314045110).

ABCD 1 1 1 1

0.8 0.8 0.8 0.8

0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4 GREMLIN

0.2 0.2 0.2 0.2 GREMLIN(no prior) GREMLIN(no prior) GREMLIN(no prior)

0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 DCA PSICOV MIc GREMLIN(no prior)

EFG1 1

0.8 0.8 0.8

GREMLIN 0.6 0.6 0.6 GREMLIN (no prior) PSICOV DCA 0.4 0.4 MIc 0.4 Accuracy Accuracy at L/2 0.2 Accuracy at L/5 0.2 0.2 Fraction of Targets

L/2 L 3L/2 1 5 10 20 1 5 10 20 Number of Predictions Sequences Per Position Sequences Per Position

Fig. 1. Accuracy of contact prediction. Comparison of GREMLIN with DCA (A), PSICOV (B), MIc (C), and GREMLIN when prior information is used (D). Each point corresponds to a protein, the axes indicate the accuracy of the top ranked L/2 Cβ−Cβ contacts predicted by the indicated methods. (E) (solid lines) Average accuracy for varying numbers of predictions; (broken lines) fraction of targets where GREMLIN was more accurate than the indicated method. Dependence of accuracy of the top L/5 (F) and L/2 (G) predictions on the alignment depth for a subset of 75 targets with deep alignments.

www.pnas.org/cgi/doi/10.1073/pnas.1319550110

18734–18735 | PNAS | November 12, 2013 | vol. 110 | no. 46 www.pnas.org Downloaded by guest on September 27, 2021 NEUROSCIENCE Correction for “Trafficking of gap junction channels at a verte- Sci USA (109:E573–E582; first published February 7, 2012; brate electrical synapse in vivo,” by Carmen E. Flores, Srikant 10.1073/pnas.1121557109). Nannapaneni, Kimberly G. V. Davidson, Thomas Yasumura, The authors note that the legend for Fig. 4 appeared incorrectly. Michael V. L. Bennett, John E. Rash, and Alberto E. Pereda, The figure and its corrected legend appear below. which appeared in issue 9, February 28, 2012, of Proc Natl Acad

Fig. 4. Presence of endocytic and exocytic machinery in CEs. CEs are identified by immunofluorescence connexin labeling with monoclonal anti-Cx35/36 (mCx35) (green). (A) Diagram of the M-cell. (Inset) Confocal projection of a single CE (from Fig. 4D of ref. 23; this false-color image was rotated and recolored for consistency with adjacent immunofluorescence). Single-terminal images in this figure represent the average of three to five z-sections; the dotted line denotes the perimeter of a single CE. (B) Confocal projection of a portion of the lateral dendrite of the M-cell using double immunolabeling with polyclonal anti-Cx36 (pCx36; Zymed 36-4600; green) and monoclonal anti-SNAP-25 (mSNAP-25; red) antibodies. (C and D) Confocal projection of single CEs using double immunolabeling with anti-Cx36 (green) and monoclonal anti–SNAP-25 (red) antibodies. In en face view (C), SNAP-25 was not restricted to the periphery of the terminals, where glutamate receptors and active zones are concentrated, but also was found in the central region where GJs predominate (arrowheads).(E and F) As with SNAP-25, some dynamin labeling (monoclonal antibody) was closely associated with labeling with anti-Cx36 (arrowheads), consistent with endocytosis of cell–cell channels. The image in F is a tilted side view of a CE, whose contact area is characteristically concave with the center protruding into the M-cell. The red dotted line in F indicates the approximate position of the cell surface (“M-cell” indicates the dendritic side). The approach does not distinguish between presynaptic and postsynaptic locations of connexin labeling. Both dynamin and SNAP-25 labeling also were observed in the vicinity of CEs (asterisks in D and E). Most of these sites were anti-Cx36 negative and likely correspond to small inhibitory boutons (65).

www.pnas.org/cgi/doi/10.1073/pnas.1319624110 CORRECTIONS

PNAS | November 12, 2013 | vol. 110 | no. 46 | 18735 Downloaded by guest on September 27, 2021 Trafﬁcking of gap junction channels at a vertebrate PNAS PLUS electrical synapse in vivo

Carmen E. Floresa, Srikant Nannapanenia, Kimberly G. V. Davidsonb, Thomas Yasumurab, Michael V. L. Bennetta,1, John E. Rashb,c, and Alberto E. Peredaa,1 aDominick P. Purpura Department of Neuroscience, Albert Einstein College of Medicine, Bronx, NY 10461; bDepartment of Biomedical Sciences, Colorado State University, Fort Collins, CO 80523; and cProgram in Cell and Molecular Biology, Colorado State University, Fort Collins, CO 80523

Contributed by Michael V. L. Bennett, December 30, 2011 (sent for review December 22, 2011) Trafficking and turnover of transmitter receptors required to main- connexons are removed from the center of the plaques (15–18) as tain and modify the strength of chemical synapses have been char- intact regions of the junction that are internalized into one or the acterized extensively. In contrast, little is known regarding trafficking other of the coupled cells, with larger internalized areas forming of gap junction components at electrical synapses. By combining double-membrane vesicles that in thin sections can appear as an- ultrastructural and in vivo physiological analysis at identified mixed nular GJs (asterisk in Fig. 1B). These structures subsequently are (electrical and chemical) synapses on the goldfish Mauthner cell, we transported to lysosomes, where they are degraded (17, 19). show here that gap junction hemichannels are added at the edges of However, whether these processes occur in native neuronal GJs in GJ plaques where they dock with hemichannels in the apposed vivo remains unknown. membrane to form cell–cell channels and, simultaneously, that intact We investigated trafficking of GJ channels and the functional junctional regions are removed from centers of these plaques into relevance of this trafficking at identifiable mixed (electrical and either presynaptic axon or postsynaptic dendrite. Moreover, electri- chemical) synapses on the goldfish Mauthner (M-) cell. These cal coupling is readily modified by intradendritic application of pep- synapses are the large myelinated club endings or simply club tides that interfere with endocytosis or exocytosis, suggesting that endings (CEs) (Fig. 1A) (20). CEs are advantageous for correla- the strength of electrical synapses at these terminals is sustained, at tions of their structural and biochemical features with their in vivo least in part, by fast (in minutes) turnover of gap junction channels. A physiological properties and provide a valuable model for the study peptide corresponding to a region of the carboxy terminus that is of vertebrate electrical transmission (21, 22). CEs are unusually NEUROSCIENCE conserved in Cx36 and its two teleost homologs appears to interfere large (5–10 μm in diameter), and each has ∼100 GJ plaques and up with formation of new gap junction channels, presumably by reduc- to ∼100,000 channels, containing connexin 35 (Cx35) (23), a fish ing insertion of hemichannels on the dendritic side. Thus, our data ortholog of the mammalian neuronal connexin, Cx36 (24). Recent indicate that electrical synapses are dynamic structures and that their data indicate that a second homolog of Cx36, Cx34.7, also is channels are turned over actively, suggesting that regulated traffick- present at these terminals (25). Both electrical and chemical ing of connexons may contribute to the modification of gap junc- components of transmission can undergo long-term potentiation tional conductance. or depression as a result of presynaptic impulse activity (26–30). Consistent with these dynamic properties, we found ultrastructural connexin | synaptic plasticity | auditory | potentiation | freeze-fracture evidence for active insertion of hemichannels and removal of cell– cell channels at these GJs that may contribute to the dynamic onstitutive and regulated trafficking of ion channels to and properties. We found that electrical transmission was strengthened Cfrom the plasma membrane are important processes for the in minutes by intradendritic application of peptides that interfere maintenance of cell function and responsiveness to environmental with endocytosis and weakened by peptides that reduce insertion stimuli. Trafficking of ionotropic receptors involving processes of of new hemichannels, indicating fast turnover of GJ channels. insertion and retrieval maintains the strength of chemical synapses Furthermore, intradendritic application of a peptide correspond- and underlies several forms of synaptic plasticity (1–4). Trafficking ing to the last 15 amino acids of the CT of Cx36, including a PDZ- usually requires interactions between the receptor’s carboxy ter- binding domain (31) (a region that is highly conserved between fi minus (CT) and a variety of cytosolic proteins, including those Cx36 and its sh homologs), progressively depressed electrical forming scaffolds (3, 5–9). In contrast, little is known regarding the transmission, suggesting that this portion of the molecule con- involvement of channel trafficking in modulating the strength of tributes to targeting and/or insertion of new GJ hemichannels. gap junction (GJ)-mediated electrical transmission. Taken together, our data indicate the existence of active turnover Vertebrate GJs are clusters (plaques) of aqueous integral of GJ channels in vivo; this turnover may be modulated to increase membrane protein channels that connect the interiors of the or decrease synaptic strength. coupled cells. Each channel is formed by the docking of two hex- Results americ connexin hemichannels, or connexons, one contributed by Ultrastructural Evidence for Trafficking of GJ Channels. Freeze- each of the adjoining cells, forming molecular pathways for the fi direct transfer of signaling molecules and for the spread of elec- fracture immunogold-labeled (FRIL) replicas of gold sh hind- trical currents between cells. Although many neuroscientists have brain (23) were examined for ultrastructural evidence of turnover of GJ channels in CEs (n = 20 from three dendrites, two considered electrical synapses to be passive intercellular channels fi fi that lack modifiability, they are proving to be dynamic structures sh) (Table 1). These large terminals were identi ed using fi confocal grid mapping of M-cells that had been injected with having distinct pathways for synthesis, traf cking, and degradation fi (10–13) and subject to posttranslational modifications such as Lucifer yellow during in vivo recordings before tissue xation phosphorylation (14). Trafficking and turnover of recombinant GJ channels have been observed in cell-expression systems, where – Author contributions: C.E.F., S.N., J.E.R., and A.E.P. designed research; C.E.F., S.N., K.G.V.D., they have been investigated extensively (15 18). A wealth of evi- T.Y., J.E.R., and A.E.P. performed research; C.E.F., S.N., K.G.V.D., M.V.L.B., J.E.R., and dence indicates that new connexins are trafficked in vesicles as A.E.P. analyzed data; and C.E.F., M.V.L.B., J.E.R., and A.E.P. wrote the paper. undocked hemichannels that are inserted into the plasma mem- The authors declare no conflict of interest. brane at the periphery of existing GJ plaques (Fig. 1B), where they 1To whom correspondence may be addressed. E-mail: [email protected] or dock with hemichannels in the apposed membrane (16, 18). Older [email protected]. www.pnas.org/cgi/doi/10.1073/pnas.1121557109 PNAS Early Edition | 1of10 Fig. 1. Neuronal GJs at large myelinated CEs; evidence for insertion of connexin hemichannels. (A) CEs are large, mixed (electrical and chemical) synaptic contacts on the lateral dendrite of the M-cell. (B) Trafficking and turnover of recombinant GJ channels. New connexons are trafficked to the membrane in vesicles as unpaired hemichannels, where they are inserted at the periphery of the GJ plaque (red) and dock with hemichannels in the apposed membrane. Subsequently, they are internalized as small clusters of entire channels (green) into either of the coupled cells from regions near the center of the plaque. (C–J) Ultrastructural evidence for insertion of GJ hemichannels in pre- and postsynaptic membranes at CEs (stereoscopic images). White boxes in C are enlarged in D–G as indicated by numbers within boxes. (C)TenGJs show intermixing of larger and smaller plaques, and there are eight presumptive exocytotic depressions at various stages of insertion into the M-cell dendritic surface. These depressions are typical of membrane exocytosis and contain about one dozen 9-nm IMPs; most are adjacent to larger GJ plaques. Both P-face IMPs (C and D) and E-face pits (J) are in crystalline hexagonal array throughout most of each GJ. (D) At higher magnifica- tion, one exocytotic depression (#5) appears to be in a late stage of exocytosis. This image also shows two shallow depressions (green overlays) within the GJ plaque, apparently representing channel removal (Fig. 2). (D–I) New connexin hemichannels apparently are trafficked in vesicles and inserted into the plasma membrane at or near the periphery of GJ plaques. The final stage is shown in J, where clusters of irregularly distributed connexon pits (red overlays) appear to reflect ongoing incorporation into the edge of GJ plaque. (C–G)Five proposed stages in exocytosis of hemichannel-containing vesicles, all from the same synapse. Hemichannel-size (9-nm) IMPs appear in collapsing exocytotic depressions (red overlay). Recently fused vesicles (stages #1–3) have hemichannel IMPs deep within the vesicles, whereas flattening vesicles have IMPs escaping from the vesicle remnant margins and apparently migrating to the periphery of the nearest GJ (stages #4–5). Because of the low level of Cx34.7 labeling (using Ab JOB 2930/1) in this replica, #811, fusing exocytotic vesicle remnants were labeled only rarely by immunogold. (H and I) Immuno- gold labeling for Cx34.7 (arrowheads) associated with 9- nm IMPs in exocytotic pits in an M-cell near the periphery of existing GJs (from DRD#1). (J) Immunogold beads labeling connexins in the apposed membrane beneath the clustered E-face pits at the edge of a GJ (red overlays) is consistent with docking of connexons to the labeled hemichannels in the apposed membrane and therefore the coordinated formation of cell–cell channels that are not yet in crystalline array (Ab298, replica #812; for technical reasons, the level of labeling for Cx35 is higher in this replica than for Cx34.7 in replica #811). Abbrevia- tions for Figs. 1 and 2: CE-E, CE E-face; CE-P, CE P-face; MC- E, M-cell E-face; MC-P, M-cell P-face. (Scale bars, 0.1 μm.)

(Fig. 1; see Experimental Procedures for details). GJs were seen as and CE. Examples of deformations in M-cell P-faces are illus- aggregates of intramembrane particles (IMPs) in the pro- trated in Fig. 1 C–I. These cup-like deformations did not resemble toplasmic leaflet (P)-face or as aggregates of pits in the extrap- exocytosis of synaptic vesicles, which occurs only presynaptically lasmic leaflet (E)-face, both of which were largely in hexagonal and is localized to active zones. Moreover, they are unlikely to arrays. As previously reported ( 23, 32), the size of GJ plaques in correspond to postsynaptic insertion of glutamate receptors, which CEs was highly variable, with large and small plaques in- are seen as distinctive E-face rather than P-face IMPs (33) and terspersed. Immediately adjacent to approximately half of the were not labeled by the connexin antibodies. large plaques (>200 connexons), we found abundant membrane The P-face deformations appear to correspond to various deformations that were consistent with insertion of GJ hemi- stages of membrane insertion, from small cup-like invaginations channels; these deformations were hemispherical to almost containing about one dozen 9-nm P-face IMPs (Fig. 1E, #1)to flattened membrane patches in the P- and E-faces of the M-cell nearly flattened invaginations (Fig. 1F, #2–4). These sites of

2of10 | www.pnas.org/cgi/doi/10.1073/pnas.1121557109 Flores et al. Table 1. Number of synapses examined in three replicas and and the yellow arrow points to upward deformation into the PNAS PLUS the number of GJs, insertion profiles, and retrieval profiles in overlying CE). These small invaginations within the central those synapses regions of GJs were about four times as abundant toward the M- FRIL Active/total Insertion Removal cell cytoplasm (which has abundant lysosomes) as toward the CE no. synapses No. of GJs profiles profiles cytoplasm. One large invagination that was cross-fractured and of indeterminate depth extended into the CE cytoplasm (Fig. 2C, 811 4/5 174 22 18 aqua overlay). In the cross-fractured invagination, M-cell cyto- 812 6/6 153 15 39 plasm can be seen (asterisk in Fig. 2C). By FRIL, we did not DRD1 5/9 179 8 11 detect detached annular GJs within the cytoplasm of either cell. Totals 15/20 506 45 68 In both large and small invaginations into both M-cell and CE, the fracture plane occasionally stepped from E-face to P-face All removal profiles were within large GJs (>200 connexons), and most of the insertion profiles were associated with the large GJs. In contrast, few, if within the margin of the invagination, revealing P-face IMPs of any, of the insertion profiles were uniquely associated with small GJs (<50 the underlying cell membrane and E-face pits of the overlying connexons) (i.e., there always was a nearby medium or large GJ). cell (Fig. 2 B, blue arrow, and C, black arrows). Such images confirm that both pre- and postsynaptic membranes were simultaneously invaginated into either CE or M-cell cytoplasm apparent vesicle insertion near GJs often were labeled either and that these endocytotic membrane vesicles contained intact with anti-connexin antibody JOB 2930/1 (Fig. 1 C–I) or with (docked) channels linking the two apposed membranes (Fig. 2 B ab298 (Fig. 1J). (See Experimental Procedures for antibodies and C, arrows). In any case, irregular distributions of IMPs/pits, used.) Moreover, IMPs at putative insertion sites often appeared usually within shallow invaginations, may represent early mark- to be in transit to a nearby GJ (Fig. 1 D, #5; F,#2–3; and G, ers for membrane endocytosis. Moreover, the irregularly arrayed #4). IMPs/pits that were between the exocytotic profiles and the IMPs/pits in endocytosing membranes (Figs. 1D and 2 B and C; edge of the nearest GJ were not yet in crystalline array; never- green overlays) imply physical stresses sufficiently strong to dis- theless, the hemichannels apparently were linked structurally to rupt the normal crystalline organization of GJ channels during hemichannels in the apposed membrane (Fig. 1J; red overlays), endocytotic deformation. Finally, many otherwise crystalline GJs as suggested by immunogold labeling of the irregularly distrib- had central areas that were devoid of connexon IMPs/pits, uted E-face pits at the edges of GJs. Because GJ E-face pits have leaving in each a smooth, particle-free area of intervening GJ no connexin protein to label (34, 35), the immunogold beads membrane (Fig. 2A, Upper Right and Lower Left Insets). If it is NEUROSCIENCE beneath peripheral irregularly arrayed E-face pits must represent assumed that the peripheral connexins in larger plaques are labeling of hemichannels in the subjacent membrane (as illus- locked rigidly into the normally regular hexagonal array, removal trated in Fig. 6). Indeed, coordinated docking of isolated pairs of of 20–40 connexons from the center of a rigid plaque initially hemichannels, one in each apposed membrane, has been re- would require lateral diffusion of lipids to fill the membrane void ported in developing GJ formation plaques in vitro (36). In our created by endocytosis—a possibility that seems to be supported in vivo material, intercellular linkage of the dispersed hemi- by images of the central areas of larger GJs that are completely channels is suggested by the increased shadow length of the devoid of connexon particles/pits (Fig. 2A, Upper Right Inset). dispersed IMPs adjacent to the GJs (Fig. 1G). This feature is These clear areas presumably would represent a late stage of GJ consistent with taller IMPs that form linked hemichannels in remodeling, with the final stage represented by migration of partially narrowed extracellular space like that at GJ formation channels into the centers of those plaques and reformation of plaques in vitro (36). the crystallization array. Alternatively, loose arrays of particles in Labeling of the dispersed clusters of E-face pits at the pe- the central areas may be an indicator of channels moving into the riphery of GJs (Fig. 1J), where no protein is present in the empty region to form new regions of crystalline arrays. replicated membrane face (see Fig. 6), indicates that multiple In summary, our ultrastructural data suggest that GJ hemi- hemichannels are immediately beneath the E-face pits, consis- channels in CE synapses are delivered to the apposed plasma tent with coordinated insertion of connexon-containing vesicles membranes by intercellularly coordinated fusion of small, uni- at approximately mirror locations at the periphery of individual form-size exocytotic vesicles, each containing about a dozen GJs linking the two coupled cells (Fig. 1J). Alternatively, such hemichannels, and that these hemichannels are directed to the labeling could imply a mechanism for directed migration of free nearest GJ rather than drifting randomly in the plasma mem- connexons to a limited number of peripheral docking sites. No- brane. Moreover, the few areas of irregularly distributed E-face tably, the paired irregular clusters of labeled connexons beneath connexon pits at the periphery of the GJ plaques were labeled, E-face pits at the periphery of GJs were always flat and did not and IMPs at these locations were taller than those within GJ resemble any stage of membrane deformation for channel en- plaques; these observations imply docking of connexons before docytosis (as described in the next paragraph). their incorporation into the regular hexagonal crystalline arrays. In these synapses undergoing substantial membrane insertion, Consistent with in vitro observations of recombinant connexins, many of the larger GJ plaques also contained IMPs/pits that GJ channels in CE synapses appear to be removed from the were hexagonally arrayed at their peripheries but irregularly center of the GJ plaques as intact channels within regions of arrayed and reduced in packing density at their centers (Figs. 1D paired junctional membrane, with endocytosis of double-mem- and 2 A–C, green overlays). Some of these plaques had multiple brane vesicles into each of the coupled cells (16, 19, 37) but with patches of disorganized IMPs/pits separated by intervening approximately fourfold greater internalization into the M-cell. crystalline IMPs/pits, suggestive of several discrete areas of membrane endocytosis. (Fig. 1D shows two such shallow inva- Interference with Endocytosis and Exocytosis Modifies Synaptic ginations.) These irregular regions ranged from slightly concave Strength. The coexistence of ultrastructural features consistent or convex [depending on whether viewed from outside or from with both insertion and removal of GJ channels within the same within the endocytosing M-cell (Fig. 1D) or CE] to deeply con- CE synapse suggests relatively continuous turnover of channels cave/convex into the M-cell (Fig. 2B) vs. into the CE (Fig. 2C), at these synapses. To test this hypothesis further, we used pep- respectively. In some cases, membrane deformations were seen tides (released by diffusion from an intradendritic recording pi- projecting into both cells, representing simultaneous double- pette) to interfere with endocytosis or exocytosis. We measured membrane endocytosis into both cells (Fig. 2B, green overlays; the peptides’ effects on the electrical (GJ-mediated) and gluta- the blue arrow points to deformation into the underlying M-cell, matergic components of the mixed excitatory postsynaptic po-

Flores et al. PNAS Early Edition | 3of10 Fig. 2. Evidence for endocytotic removal of GJ channels. (A) Thirteen GJs (blue overlays) including five larger GJ plaques, have six regions of noncrystalline connexons (green overlays), apparently reflecting early or late stages of endocytosis into the M-cell and CE. Upper Right Inset (yellow box) contains a relatively large patch of IMP-free M-cell E-face in a GJ from a more distant region of the same replica. The connexon-free patch may represent the final stage of endocytosis, occurring immediately before IMP redistribution to reestablish the normal crystalline configuration of IMPs/pits seen in synapses not undergoing endocytosis. Note that the clear patch has three E-face IMPs of unknown composition. Lower Left Inset shows an enlargement of a small patch of IMP-free membrane at the center of the adjacent GJ, presumably representing ongoing recrystallization of cell–cell channels. In the boxed region (labeled “B” in A and enlarged as stereoscopic images in B), endocytosis appears to be into both the M-cell (blue arrow) and the CE (yellow arrow). The noncrystalline E-face pits within the endocytotic profiles (and a few noncrystalline P-face particle arrays at the invagination into the M-cell) are distinguished from the surrounding hexagonally arranged connexons (uncolored regions). (C) Stereoscopic views: During internalization of a larger patch of GJ membrane, the invagination consists of two closely apposed membranes, forming a double-membrane structure similar to that reported in cell-expression systems (19). Asterisk indicates M-cell cytoplasm (aqua overlay) being endocytosed into the CE. Arrows indicate P-face IMPs aligning with E-face pits. A–C are from replica #812; connexin ab298. (Scale bars, 0.1 μm.) tential (mixed EPSP) evoked by stimulation of the VIIIth nerve. of the synaptic potential are separated in time and can be iden- Because of the brief membrane time constant of the M-cell (∼0.4 tified unambiguously and measured reliably (Fig. 3A). Intra- ms) and duration of the VIIIth nerve volley, the two components dendritic application of the dynamin-inhibitory (D-15) peptide

4of10 | www.pnas.org/cgi/doi/10.1073/pnas.1121557109 Flores et al. PNAS PLUS NEUROSCIENCE

Fig. 3. Interfering with endocytosis and exocytosis modiﬁes synaptic strength. (A) Experimental arrangement used to test the effect of intradendritic application of interfering peptides in the M-cell (Left) on the electrical and glutamatergic components of the mixed EPSP evoked by stimulation of the VIIIth nerve and CE activation (Right). (B) D-15 caused a run-up of the electrical component (black circles), which was not observed with a scrambled version of this peptide (open circles). (C) SNAP-25 inhibitory peptide (black circles) caused a progressive run-down of electrical transmission, which was not observed with the vehicle solution (open circles). (D and E) D-15 peptide increased the glutamatergic component of the mixed EPSP (D), whereas SNAP-25 decreased it (E). The control responses were taken from the same records as the controls in B and C.

(amino acids 828–842 of rat dynamin) (Fig. 3B), a proline-rich differential susceptibility to the action of cytosolic peptidases domain that is thought to reduce endocytosis by interrupting and/or progressive diffusion from the application site. dynamin’s interaction with amphiphysin (38), enhanced the D-15 and SNAP-25 peptides are known to modify the strength electrical component of the mixed EPSP (172.7% ± 22.2%, n = of glutamatergic transmission at mammalian synapses (1, 4, 40– 5) (Fig. 3B, filled circles). (Here and elsewhere changes 20–30 42), and the actions on the chemical component of the mixed min after onset of recording are expressed as percent of initial EPSP provided a valuable control for their efficacy in the M-cell. value.) In contrast, no changes were observed with a scrambled Accordingly, we found that the glutamatergic component of the ± < version of this peptide (108.1% 5.9%, n =6,P 0.05 vs. mixed EPSP was enhanced within minutes by intradendritic peptide) (Fig. 3B, open circles). Conversely, intradendritic ap- ± – applications of the D-15 peptide (159.3% 29.8% at 20 min, vs. plication of a SNAP-25 peptide (amino acids 182 192 of SNAP- scrambled peptide: 103.1% ± 7.4%, n =5;P < 0.05) (Fig. 3D), 25) that interferes with formation of SNARE complexes (39) whereas it was depressed over the same time course by the SNAP- depressed the amplitude of the electrical component of the 25 peptide (80.8% ± 3.5%, vs. vehicle: 100.9% ± 4.7%, n =5;P < mixed EPSP over the same time course (to 83% ± 4.9% of initial value, n = 6) (Fig. 3C). However, the electrical component 0.05; n = 6) (Fig. 3E). The time course and magnitude of these remained unaffected during control experiments in which the changes in glutamatergic transmission were similar to those in the electrode contained only the vehicle solution (113.4% ± 7.7%, electrical component. Finally, the amplitude of the M-cell anti- ’ n =4,P < 0.01 vs. peptide) (Fig. 3C, open circles). Each peptide dromic spike, an indicator of the cell s input resistance, was not affected the amplitude of the electrical component of the mixed affected significantly during these manipulations, averaging 93.7% EPSP within minutes and with a similar time course. Differences ± 4.7% of control for D-15, 98.9% ± 6.5% for the scrambled in the magnitude of the up- and down-regulation are ascribable peptide, 92.4% ± 2.9% for the SNAP-25 peptide, and 97.9% ± to differences in the individual blocking efficacy of the peptides 7.2% for the vehicle experiments (P > 0.1 for each comparison). at the concentrations used, which also might be influenced by Thus, changes in the M-cell input resistance did not account for

Flores et al. PNAS Early Edition | 5of10 the (bidirectional) changes in both electrical and chemical com- of the M-cell by immunolabeling with anti-Cx35 (23, 31, 43), be- ponents of the EPSPs. cause the ∼100 GJ plaques present at each of these unusually large To validate further the actions of the peptides, we used immu- contacts are distributed throughout the appositional surface (Fig. nolabeling to examine the presence of both dynamin and SNAP-25 4A). Double-immunolabeling for Cx36 and dynamin or SNAP-25 in CEs. CEs can be identiﬁed unambiguously in the lateral dendrite revealed both trafﬁcking proteins in CEs. In high-resolution confocal images, dynamin (Fig. 4 B and C) and SNAP-25 (Fig. 4 D and E) were seen both in the center of the contacts, where the GJs predominate, and in the periphery, where the glutamatergic sites are located (Fig. 4 C–E; arrowheads indicate sites of superposi- tion). Thus, the distribution of dynamin and SNAP-25 immunore- activity in CEs is consistent with the proposed effects of the blocking peptides on electrical and chemical components.

CT of Fish Homologs of Cx36 Is Implicated in Trafficking of GJ Channels at CEs. Taken together, the effects of dynamin and SNAP-25–blocking peptides suggests a direct action on continuous turnover of GJ channels. Alternatively, interference with dynamin and SNAP might affect the surface expression of other proteins that, in turn, could modify synaptic transmission. It has been suggested that the last amino acids of the CT of Cx36 are necessary for the correct surface expression of these channels (44). We hypothesized that, if GJ channels are added rapidly by exocytosis, interference with this region of the molecule will, like the SNAP-25–blocking peptide, block connexon insertion and, as a consequence, result in a rundown of electrical transmission. To examine trafficking further, we used a peptide corresponding to the last 15 amino acids of the CT of Cx36 (CT-peptide), a region that is identical to the CT of one fish homolog, Cx35, and has only minor differences from the CT of another, Cx34.7 (Fig. 5A) (31). The peptide (1 mM) was included in the recording electrode solution, and both components of the EPSP were monitored. Intradendritic application of the CT-peptide caused a progressive reduction in the amplitude of the electrical component of the EPSP (to 79.2% ± 4.6% of control, n =5,at40 min) (Fig. 5B). In contrast, a scrambled version of the peptide did not cause reduction of the electrical component over the same time course (114.6% ± 3.9%, n =5;P < 0.005 vs. CT- peptide) (Fig. 5C). The increase in amplitude of the electrical component is not likely to be related to nonspecific effects of this scrambled peptide because it also was observed in this experimental series with electrodes containing only the vehicle solution (Fig. 3C). The M-cell antidromic spike amplitude was little affected (91.3% ± 5.95% of control for CT-peptide and 92.1% ± 10% for the scrambled version; P > 0.1), indicating that changes were not caused by a decrease in the cell’s input resistance. Interactions between electrical and chemical synapses are known to take place at CEs (28–30, 45). Although the peptide reduced overall chemical transmission (to 80.7% ± 8.6% of control at 40 Fig. 4. Presence of endocytic and exocytic machinery in CEs. (A) CEs are iden- min, compared with scrambled peptide: 122.8% ± 4.7%, P < tified by immunofluorescent connexin labeling (green). (Inset) Confocal pro- 0.05, n = 5), its effect was restricted to the electrical component jection of a single CE obtained with monoclonal anti-Cx35 (mCx35). Single- in three of these experiments (Fig. 5B), suggesting that the terminal images here and elsewhere represent the average of three to five z- peptide primarily affects electrical transmission. These results sections; the dotted line denotes the perimeter of a single CE. (B)Confocal suggest that the CT peptide affects molecular interactions be- projection of a portion of the lateral dendrite of the M-cell using double tween the CT of neuronal connexins and trafficking or secretory immunolabeling with polyclonal anti-Cx36 (pCx36; Zymed 36–4600; green) and monoclonal anti–SNAP-25 (mSNAP-25; red) antibodies. (C and D)Confocal proteins, altering the balance between membrane insertion and projection of single CEs using double immunolabeling with anti-Cx36 (green) removal. – en face C and monoclonal anti SNAP-25 (red) antibodies. In an view ( ), SNAP-25 Discussion was not restricted to the periphery of the terminals, where glutamate receptors and active zones are concentrated, but also was found in the central region, Our electrophysiological evidence suggests that the average where GJs predominate (arrowheads). (E and F) Aswith SNAP-25, some dynamin lifetime of GJ channels at CEs is on the order of tens of minutes. labeling (monoclonal antibody) was closely associated with labeling with anti- Despite the multiplicity of actions of dynamin and SNAP-25 and Cx36 (arrowheads), consistent with endocytosis of cell–cell channels. The image possible off-target effects of peptides, our results indicate the F in is a tilted side view of a CE, whose contact area is characteristically concave existence of active trafficking of GJ channels at these synapses. with the center protruding into the M-cell. The red dotted line in F indicates the approximate position of the cell surface (“M-cell” indicates the dendritic side). This conclusion is supported by ultrastructural evidence which, consistent with data from cell-expression systems for recombi- The approach does not distinguish between presynaptic and postsynaptic – locations of connexin labeling. Both dynamin and SNAP-25 labeling also were nant GJ channels (16 18), suggests that channels are added and observed in the vicinity of CEs (asterisks in D and E). Most of these sites were anti- removed from GJs simultaneously. These observations are con- Cx36 negative and are likely correspond to small inhibitory boutons (65). sistent with current views (18, 19) that exocytotic incorporation

6of10 | www.pnas.org/cgi/doi/10.1073/pnas.1121557109 Flores et al. PNAS PLUS

Fig. 5. Intradendritic application of the Cx35 CT-peptide reduced electrical transmission. (A) Sequence alignments of the last 22 amino acids of the CT of human (h) and mouse (m) Cx36 and the teleost homologs, perch (p) Cx35 and Cx34.7 show near identity. (B) The intradendritic application of a peptide

corresponding to the last 15 amino acids of the CT of Cx35 (CT-peptide) resulted in a reduction in the amplitude of the electrical but not the chemical NEUROSCIENCE component of the mixed EPSP (single experiment). (C) The progressive reduction in electrical transmission was not observed when a scrambled version of the peptide was injected (Upper: single experiment; Lower: average of five fish). (D) Electrical synapses at CEs show continued turnover, involving insertion of hemichannels, which dock with hemichannels from the other cell to form GJ channels, and internalization of aggregates of the entire channels. An increase in GJ conductance may involve an increase in insertion or docking of channels or a decrease in internalization; a decrease in GJ conductance may involve the converse changes. of new connexons occurs via unpaired hemichannels added involve the regulation of their insertion, their removal, or both. outside the margins of preexisting GJs, whereas connexons are A simple possibility is that rate of insertion of channels is con- removed from the center of GJ plaques as clusters of intact cell– stant and that the strength of electrical transmission is regulated cell channels in double-membrane vesicles. by the modification of the rate of removal of channels from Because of their susceptibility to the action of peptidases and preexisting GJ plaques. Changes in the rate of removal could diffusion from the application site, peptides are not ideal tools for either increase or decrease the number of functional intercellular accurate estimations of the turnover rate of GJ channels at CEs. channels in the GJ plaque to establish a new balance. Alterna- Despite these considerations, estimates of turnover time from tively, variations in rate of insertion could modulate amplitude in experiments with the SNAP-25 and CT-peptides indicate an overall the face of a fixed rate (or fixed rate constant) of removal. In half-life of ∼1–3 h, a value that is consistent with previous estimates addition, a set point also could depend on the properties and for turnover of GJs in expression systems and intact tissue (46, 47) composition of the scaffold and regulatory proteins that sustain and for other membrane channels and receptors (47). This rapid channel function in GJs and that underlie the trafficking and turnover is in contrast to GJs in adult lens cells (adult lens cells lack stabilization of GJ channels. nuclei and protein synthetic machinery; thus their proteins, in- Alternative mechanisms of modulation of GJ conductance cluding connexins, must last a lifetime) (48). Our results also in- exist. Cx36 and its two fish homologs are targets of regulation by dicate that postsynaptic modifications of GJ channel trafficking can protein kinases (43, 51, 52), and direct phosphorylation of con- affect the rate at which postsynaptic hemichannels link to pre- nexins, rather than the incorporation of new channels, was pro- synaptic hemichannels to form functional channels, an observation posed to underlie at least some regulation of Cx36-mediated that is consistent with the fact that potentiation of electrical electrical synapses in the retina (52, 53). Protein kinase A and transmission at CEs is induced postsynaptically (26, 28, 30, 49). CaM-KII (43, 49) mediate activity-dependent changes in the Our results show that the strength of electrical transmission at strength of electrical transmission at CEs. Furthermore, activity- CEs is maintained by a balance between hemichannel insertion, dependent changes of electrical transmission at CEs can take cell–cell channel formation, and channel removal, because pep- place within minutes (27–30) and also can be transient (27); thus, tides interfering with endocytosis and exocytosis modify within modifications can take place within a time frame faster than that minutes the amplitude of the electrical and chemical compo- of the trafficking observed here. As proposed for long-term ponents of the mixed EPSP. These findings suggest that there may tentiation of glutamatergic transmission at hippocampal synap- be a set point, so that drift away from this value activates feed- ses (55), sequential molecular mechanisms involving changes in back that modulates trafficking to maintain a more-or-less con- gating and trafficking may underlie modifications of electrical stant number of open intercellular GJ channels. Such a dynamic transmission at CEs. The initial expression of the potentiation process also could be involved in plastic modifications of junc- thus could involve posttranslational modifications of channels tional conductance (28, 29, 49, 50), and activity-dependent al- already present in the junctions, whereas the longer-term po- teration of electrical transmission (26, 30) could involve changing tentiation could be mediated by adjusting channel turnover. the set point. Modulation of the number of open channels could Increases in electrical coupling are unlikely to involve the insertion

Flores et al. PNAS Early Edition | 7of10 of recycled channels, which is a likely mechanism at glutamatergic synapses (56), and current evidence indicates that internalized GJs are degraded in lysosomes (15). Our experiments demonstrate that electrical transmission is readily modified by interfering with trafficking of GJ channels. However, early estimates obtained from the average number of GJ channels in individual CEs (59,344 cell–cell channels) (32) and an assumed single-channel conductance of 100 pS (57) suggested that only a small percentage of the morphologically observed channels were open during the electrical component of the EPSP. We know now that the single-channel conductance of Cx35 (58) and Cx36 channels (59) is lower, ∼15 pS, indicating that only 2–4% of the channels in a CE are open during a unitary electrical EPSP. Be- cause functional GJ channels characteristically have an open probability of ∼1 at a transjunctional voltage of zero (60), electrical transmission at single CEs likely results from a small number of channels with a high open probability rather than from a larger number with a low open probability. In any case, although CE synapses contain a large fraction of channels that are electrically silent, electrical transmission is susceptible to manipulations of endocytosis and exocytosis. Furthermore, because the dynamic Fig. 6. Diagram illustrating the fracture plane (yellow line) in a CE/M-cell GJ range of variation of synaptic strength of electrical synapses in CEs (A) and labeling of cytoplasmic epitopes of connexins beneath the replica (29) requires only a small fraction of the channels that are con- (B). (A) In vertebrate GJs, freeze-fracturing separates connexons at the site of tained in the contact, the striking disparity between conductive and apposition in the extracellular space, with all connexons of the upper plaque nonconductive channels is unlikely to correspond to the existence removed and all connexons of the lower plaque retained beneath the rep- of a functional reserve. However, what appears to be an over- lica. Thus, only protein of connexons of the lower replica remaining after whelming excess of nonconducting channels could be of functional SDS treatment can be immunolabeled. (B) In all cases, FRIL labels the con- relevance if connexins have an additional function, such as acting nexons in the cell whose cytoplasm was retained beneath the replica (18-nm as adhesion molecules (61, 62). Docking of connexons requires gold beads are black; primary and secondary antibodies are indicated by black “Ys”) and does not label the E-face pits from which the connexons of a specific distance between the membranes of adjoining cells, and – the upper cell were removed by fracturing, but instead labels the underlying cell cell channels that did not open might contribute to the me- connexons remaining in the attached membrane of the lower cell. Labeling chanical stability necessary to maintain functional intercellular of M-cell connexons, including in a fusing vesicle with connexon hemichannels. The generation of this striking disparity between con- channels (Left; red lipid bilayer corresponding to FRIL images in Fig. 1 C–J), ductive and nonconductive channels is likely to reflect essential the GJ plaque (Center; green lipid bilayer corresponding to overlays in FRIL aspects of GJ formation and maintenance, including the lifetime images in Fig. 2), an internalizing endocytotic double-membrane in- of channels and the existence of heterogeneous populations of vagination (Center; green lipid bilayer), and paired patches whose con- channels within the plaque. nexons are slightly more widely separated and are on the tips of deformed In summary, our results provide evidence that trafficking is lipid cones that produce slightly larger and deeper E-face pits and slightly taller P-face IMPs (Fig.1J, lower red overlay, and Fig. 1G, right side of red important in regulating the strength of transmission at electrical fi overlay, respectively), suggesting that the connexons are docked across synapses. Because traf cking necessarily involves the interaction a partially narrowed (10-nm) extracellular space, as described in ref. 36. Note of GJ channels with scaffold and regulatory proteins, our results that in the endocytosing membrane (green lipid bilayer), intact connexons suggest that electrical synapses, as proposed for other GJs (63), are spaced more widely in the outer concave layer than in the coupled inner no longer should be viewed as simple clusters of intercellular convex layer (A). (B) Pt with arrow indicates platinum shadowing direction channels but rather should be seen as complex structures in and 60° angle of deposition; black line (1–2 nm thick) indicates an electron- which many molecular elements participate in both short-term opaque platinum layer that is visualized by TEM; gray line (20 nm thick) and long-term regulation. indicates a carbon support layer that is applied by rotary thermionic deposition on top of the Pt layer. (The carbon layer is almost transparent by Experimental Procedures TEM, except on nearly vertical surfaces.) Segments corresponding to Figs. 1 E–J and 2C are labeled. Freeze-Fracture Replica and Immunogold Labeling. A combination of physiological and anatomical techniques was used to identify the VIIIth nerve Carassius auratus fi terminals on the M-cell in . After formaldehyde xation, as sequences in Cx34.7 and Cx35) (23) and counter-labeled for 13 h with the above, two specimens were processed by single-replica FRIL (23), and one same goat anti-rabbit gold-conjugated secondary antibodies used for sam- additional specimen was processed by double-replica SDS-digested fracture ple #811. For the double-replicas DRD#1 (top and bottom), both matching replica labeling (SDS-FRL) (34, 65). Slices containing the lateral dendrite of replica complements were mounted on Lexan-coated index grids, labeled M-cells that were injected with Lucifer Yellow during intracellular in vivo simultaneously for 4 h with rabbit polyclonal antibody against Cx34.7 (JOB recordings were freeze-fractured and coated with 1.5 nm of platinum and 2930/1) and mouse monoclonal antibody directed against Cx35 (MAB3043; 20 nm of carbon (Fig. 6). For single-replica samples, a gold index (aka Millipore Bioscience Research Reagents), and then counter-labeled with “Finder”) grid was bonded to the coated surfaces using Lexan plastic dis- 5-nm goat anti-rabbit IgG and 10-nm goat anti-mouse IgG (both from solved in dichloroethane; the samples were thawed and grid-mapped by confocal microscopy to determine the location of the M-cell lateral dendrite. BB International). Then cellular material was removed from the side opposite the grid by Electrophysiological Techniques and Data Collection. gentle washing with SDS detergent. For both single- and double-replica Intracellular in vivo fi Carassius auratus samples, residual connexin proteins adhering to the replica after SDS recordings from adult gold sh ( ) were performed as pre- washing were labeled as follows: Replica #811 (Fig. 1 C–G) was labeled for viously described (28). Responses to VIIIth nerve or spinal cord stimulation 1 h with rabbit anti-Cx34.7 (JOB 2930/1) and counter-labeled for 13 h with were recorded from the lateral dendrite of the M-cell (300–400 μm from the goat gold-conjugated anti-rabbit secondary antibodies from Chemicon (10 axon hillock) using glass microelectrodes (4–8MΩ) containing 0.5 M KCl, pH nm), Jackson ImmunoResearch (12 nm), and BB International (10 nm). Rep- 7.2 (see below). Because of the brief membrane time constant of the M-cell lica #812 was labeled for 1 h with rabbit polyclonal antibody against Cx36/ (∼400 ms), both components of the synaptic potential evoked by stimulation Cx34.7/Cx35 (Ab298 made against an 18-amino acid sequence in human Cx36 of the VIIIth nerve can be identified unambiguously and measured reliably, beginning at amino acid #298, which is almost identical to the corresponding because the brief duration of the electrical component does not significantly

8of10 | www.pnas.org/cgi/doi/10.1073/pnas.1121557109 Flores et al. influence the peak of the subsequent chemical response [peak amplitudes 1× PBS + 0.3–0.4% Triton X-100, vol/vol, pH 7.4) + 5–10% normal goat serum PNAS PLUS were detected within a range of points using Igor Pro software (Wave- (NGS). As a marker for CEs, sections were incubated overnight at 4 °C on Metrics)]. Synaptic potentials were monitored every 4 s, and traces were a tilting platform in PBSTr + NGS containing either monoclonal anti-Cx35 averaged in sets of 15 or more. Amplitude changes were estimated by (1:250/1:500) (Millipore Bioscience Research Reagents) or rabbit polyclonal comparing averages obtained at baseline and 20–40 min after each experi- anti-Cx36 (1:250/ 500) (36–4600; Zymed). In addition, samples were labeled mental manipulation. Results are expressed as mean ± SEM. Student’s t test with monoclonal anti-SNAP-25 (1:100) (SYSY-111–011; Synaptic Systems) or was used to assess statistical significance. For intradendritic applications, the anti-dynamin (1:100) (BD-610245; BD Biosciences) (incubated for only 1 h in synthetic peptides (see below), at 1–2 mM concentrations, were added to this case). In addition samples were labeled with monoclonal anti-SNAP-25 the recording vehicle solution (0.5 M KCl + 10 mM Hepes, pH 7.2). In other (1:100) (SYSY-111–011; Synaptic Systems) or anti-dynamin (1:100) (BD- experiments, a 5% solution of Lucifer yellow carbohydrazide (lithium salt; 610245; BD Biosciences) (incubated for only 1 h in this case). Sections were Molecular Probes) in distilled water was injected iontophoretically into M- rinsed in PBSTr three to four times for 10 min at each rinse and incubated for cell somata for subsequent processing for FRIL. 1–2 h at room temperature with Alexa Fluor 488-conjugated goat anti- rabbit and/or Alexa Fluor 594-conjugated goat anti-mouse secondary anti- Synthetic Peptides. The following synthetic peptides were obtained from the bodies. Sections then were rinsed with PBSTr and then with 50 mM phos- Protein Chemistry Facility of Tufts University (Boston): (i) S-15, a scrambled phate buffer, pH 7.4 and mounted on slides in a propyl gallate-based version of the D-15 peptide (PRPPGPASQPPNRPV); the D-15 peptide corre- antifading solution to reduce photobleaching. Control sections, which were sponds to an amino acid sequence (828–842) that is crucial for dynamin’s incubated with secondary antibodies in the absence of primary antibodies, interaction with amphiphysin located in the PRD domain (PPPQVPSRPNRAPPG) did not show meaningful labeling. The specificity of the anti-dynamin an- (Tocris). (ii) SNAP-25 inhibitory peptide; the amino acid sequence represents tibody was confirmed by Western blot; the anti-SNAP-25 antibody (Synaptic the CT (amino acids 182–192) of SNAP-25 except that serine 187 was changed Systems) was reported to label SNAP-25 specifically in both vertebrates to alanine. This peptide inhibits SNARE-dependent exocytosis. (iii)Cx35CT15, and invertebrates. corresponding to the last 15 CT amino acids (VPNFGRTQSSDSAYV) of Cx36/35; (iv) Cx35CTs15 (YPSANTGVQFSDVSR), a scrambled version of Cx35CT15. Confocal Microscopy and Image Processing. Sections were imaged with an Olympus BX61WI confocal microscope with a mortised fixed stage with 20× Immunohistochemistry. Fish were perfused intracardially with saline phos- air, 40× apo/340 water, and 60× oil objective lenses. FLUOVIEW FV500 phate buffer (1× PBS) at pH 7.2–7.4 for 10 min followed by cold 4% (wt/vol) software was used for data acquisition. Confocal immunofluorescence XY formaldehyde in 0.1 M phosphate buffer (paraformaldehyde; PFA) for 10 images were scanned in z-axis intervals of 0.6 μm. Z-plane sections and Z- min as previously described (31). Brains were dissected out and kept in 4% plane stacks from each image were used for image analysis with Image J PFA overnight at 4 °C. After incubation tissues were sectioned (50 μm) in 0.1 (National Institutes of Health). The unlabeled area within CEs was relatively M phosphate buffer with a TPI Vibratome (Technical Products International) small compared with labeled areas. For presentation purposes, some images for further processing. Alternatively, after perfusion, brains were transferred were processed using Adobe Photoshop (Adobe Systems) and Canvas X NEUROSCIENCE to 30% sucrose in PFA (for cryoprotection) and left until they sank; then they (ACD Systems). were mounted in Optimum Cutting Temperature compound (Tissue-Tek) μ and sectioned (50 m) with a cryostat (Leica CM3050S) for further process- ACKNOWLEDGMENTS. This research was supported by National Institutes of ing. These sections were rinsed at room temperature with PBS, blocked and Health Grants DC11099, R56DC3186, and NS552827 (to A.E.P.); S10RR5831, permeabilized for 45 min to 1 h at room temperature in PBS Triton (PBSTr, S10RR8329, NS44395, and NS44010 (to J.E.R.), and NS55363 (to M.V.L.B.).

1. Lüscher C, et al. (1999) Role of AMPA receptor cycling in synaptic transmission and 20. Bartelmez GW (1915) Mauthner’s cell and the nucleus motorius tegmenti. J Comp plasticity. Neuron 24:649–658. Neurol 25:87–128. 2. Xia J, Chung HJ, Wihler C, Huganir RL, Linden DJ (2000) Cerebellar long-term 21. Robertson JD, Bodenheimer TS, Stage DE (1963) The ultrastructure of Mauthner Cell depression requires PKC-regulated interactions between GluR2/3 and PDZ domain- synapses and nodes in goldfish brains. J Cell Biol 19:159–199. containing proteins. Neuron 28:499–510. 22. Furshpan EJ (1964) “Electrical transmission” at an excitatory synapse in a vertebrate 3. Carroll RC, Beattie EC, von Zastrow M, Malenka RC (2001) Role of AMPA receptor brain. Science 144:878–880. endocytosis in synaptic plasticity. Nat Rev Neurosci 2:315–324. 23. Pereda A, et al. (2003) Connexin35 mediates electrical transmission at mixed synapses 4. Kwon HB, Castillo PE (2008) Long-term potentiation selectively expressed by NMDA on Mauthner cells. J Neurosci 23:7489–7503. receptors at hippocampal mossy fiber synapses. Neuron 57:108–120. 24. Condorelli DF, et al. (1998) Cloning of a new gap junction gene (Cx36) highly 5. Leonard AS, Davare MA, Horne MC, Garner CC, Hell JW (1998) SAP97 is associated expressed in mammalian brain neurons. Eur J Neurosci 10:1202–1208. with the alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor GluR1 25. Rash JE, et al. (2010) Neuronal gap junctions in goldfish hindbrain are primarily or J Biol Chem – subunit. 273:19518 19524. exclusively at glutamatergic mixed synapses and are heterotypic, with Cx35 in axon 6. Dong H, et al. (1997) GRIP: A synaptic PDZ domain-containing protein that interacts terminals and Cx34.7 in somata and dendrites. Program 42.3, 2010 Neuroscience Nature – with AMPA receptors. 386:279 284. Meeting Planner. San Diego, CA: Society for Neuroscience, 2010. Available at http://www. 7. Ehlers MD (2000) Reinsertion or degradation of AMPA receptors determined by sfn.org/skins/main/pdf/abstracts/am2010/poster_presentations/PosterPresentation_SatPM. activity-dependent endocytic sorting. Neuron 28:511–525. pdf. Accessed January 18, 2012. 8. Chen L, et al. (2000) Stargazin regulates synaptic targeting of AMPA receptors by two 26. Yang XD, Korn H, Faber DS (1990) Long-term potentiation of electrotonic coupling at distinct mechanisms. Nature 408:936–943. mixed synapses. Nature 348:542–545. 9. Carroll RC, Zukin RS (2002) NMDA-receptor trafficking and targeting: Implications for 27. Pereda AE, Faber DS (1996) Activity-dependent short-term enhancement of synaptic transmission and plasticity. Trends Neurosci 25:571–577. intercellular coupling. J Neurosci 16:983–992. 10. Musil LS, Goodenough DA (1993) Multisubunit assembly of an integral plasma 28. Pereda AE, et al. (1998) Ca2+/calmodulin-dependent kinase II mediates simultaneous membrane channel protein, gap junction connexin43, occurs after exit from the ER. enhancement of gap-junctional conductance and glutamatergic transmission. Proc Cell 74:1065–1077. Natl Acad Sci USA 95:13272–13277. 11. Falk MM, Buehler LK, Kumar NM, Gilula NB (1997) Cell-free synthesis and assembly of 29. Smith M, Pereda AE (2003) Chemical synaptic activity modulates nearby electrical connexins into functional gap junction membrane channels. EMBO J 16:2703–2716. synapses. Proc Natl Acad Sci USA 100:4849–4854. 12. Diez JA, Ahmad S, Evans WH (1999) Assembly of heteromeric connexons in guinea-pig 30. Cachope R, Mackie K, Triller A, O’Brien J, Pereda AE (2007) Potentiation of electrical and liver en route to the Golgi apparatus, plasma membrane and gap junctions. Eur J chemical synaptic transmission mediated by endocannabinoids. Neuron 56:1034–1047. Biochem 262:142–148. 31. Flores CE, Li X, Bennett MV, Nagy JI, Pereda AE (2008) Interaction between 13. Evans WH, et al. (1999) Trafficking pathways leading to the formation of gap junctions. Novartis Found Symp 219:44–54; discussion 54–49. connexin35 and zonula occludens-1 and its potential role in the regulation of Proc Natl Acad Sci USA – 14. Moreno AP, Lau AF (2007) Gap junction channel gating modulated through protein electrical synapses. 105:12545 12550. phosphorylation. Prog Biophys Mol Biol 94:107–119. 32. Tuttle R, Masuko S, Nakajima Y (1986) Freeze-fracture study of the large myelinated fi 15. Laird DW (1996) The life cycle of a connexin: Gap junction formation, removal, and club ending synapse on the gold sh Mauthner cell: Special reference to the degradation. J Bioenerg Biomembr 28:311–318. quantitative analysis of gap junctions. J Comp Neurol 246:202–211. 16. Gaietta G, et al. (2002) Multicolor and electron microscopic imaging of connexin 33. Harris KM, Landis DM (1986) Membrane structure at synaptic junctions in area CA1 of trafficking. Science 296:503–507. the rat hippocampus. Neuroscience 19:857–872. 17. Laird DW (2006) Life cycle of connexins in health and disease. Biochem J 394:527–543. 34. Fujimoto K (1995) Freeze-fracture replica electron microscopy combined with SDS 18. Lauf U, et al. (2002) Dynamic trafficking and delivery of connexons to the plasma digestion for cytochemical labeling of integral membrane proteins. Application to the membrane and accretion to gap junctions in living cells. Proc Natl Acad Sci USA 99: immunogold labeling of intercellular junctional complexes. J Cell Sci 108:3443–3449. 10446–10451. 35. Rash JE, Yasumura T, Dudek FE, Nagy JI (2001) Cell-specific expression of connexins 19. Piehl M, et al. (2007) Internalization of large double-membrane intercellular vesicles and evidence of restricted gap junctional coupling between glial cells and between by a clathrin-dependent endocytic process. Mol Biol Cell 18:337–347. neurons. J Neurosci 21:1983–2000.

Flores et al. PNAS Early Edition | 9of10 36. Johnson RG, et al. (2012) Gap junction assembly: Roles for the formation plaque and 52. Kothmann WW, Massey SC, O’Brien J (2009) Dopamine-stimulated dephosphorylation regulation by the C-terminus of connexin43. Mol Biol Cell 23:71–86. of connexin 36 mediates AII amacrine cell uncoupling. J Neurosci 29:14903–14911. 37. Gumpert AM, Varco JS, Baker SM, Piehl M, Falk MM (2008) Double-membrane gap 53. Urschel S, et al. (2006) Protein kinase A-mediated phosphorylation of connexin36 in junction internalization requires the clathrin-mediated endocytic machinery. FEBS mouse retina results in decreased gap junctional communication between AII Lett – 582:2887 2892. amacrine cells. J Biol Chem 281:33163–33171. 38. Wigge P, McMahon HT (1998) The amphiphysin family of proteins and their role in 54. Kothmann WW, Li X, Burr GS, O’Brien J (2007) Connexin 35/36 is phosphorylated at Trends Neurosci – endocytosis at the synapse. 21:339 344. regulatory sites in the retina. Vis Neurosci 24:363–375. 39. Gutierrez LM, et al. (1997) A peptide that mimics the C-terminal sequence of SNAP-25 55. Lisman J, Raghavachari S (2006) A unified model of the presynaptic and postsynaptic inhibits secretory vesicle docking in chromaffin cells. J Biol Chem 272:2634–2639. changes during LTP at CA1 synapses. Sci STKE 2006:re11. 40. Lledo PM, Zhang X, Südhof TC, Malenka RC, Nicoll RA (1998) Postsynaptic membrane 56. Gomes AR, Correia SS, Carvalho AL, Duarte CB (2003) Regulation of AMPA receptor fusion and long-term potentiation. Science 279:399–403. activity, synaptic targeting and recycling: Role in synaptic plasticity. Neurochem Res 41. Carroll RC, et al. (1999) Dynamin-dependent endocytosis of ionotropic glutamate – receptors. Proc Natl Acad Sci USA 96:14112–14117. 28:1459 1473. 42. Lau CG, et al. (2010) SNAP-25 is a target of protein kinase C phosphorylation critical to 57. Lin JW, Faber DS (1988) Synaptic transmission mediated by single club endings on the fi NMDA receptor trafficking. J Neurosci 30:242–254. gold sh Mauthner cell. I. Characteristics of electrotonic and chemical postsynaptic 43. Flores CE, et al. (2010) Variability of distribution of Ca(2+)/calmodulin-dependent potentials. J Neurosci 8:1302–1312. kinase II at mixed synapses on the Mauthner cell: Colocalization and association with 58. Valiunas V, et al. (2004) Biophysical characterization of zebrafish connexin35 connexin 35. J Neurosci 30:9488–9499. hemichannels. Am J Physiol Cell Physiol 287:C1596–C1604. 44. Helbig I, et al. (2010) In vivo evidence for the involvement of the carboxy terminal 59. Srinivas M, et al. (1999) Functional properties of channels formed by the neuronal gap domain in assembling connexin 36 at the electrical synapse. Mol Cell Neurosci 45: junction protein connexin36. J Neurosci 19:9848–9855. 47–58. 60. Srinivas M, et al. (1999) Voltage dependence of macroscopic and unitary currents of 45. Pereda AE, Rash JE, Nagy JI, Bennett MV (2004) Dynamics of electrical transmission at gap junction channels formed by mouse connexin50 expressed in rat neuroblastoma club endings on the Mauthner cells. Brain Res Brain Res Rev 47:227–244. cells. J Physiol 517:673–689. 46. Beardslee MA, Laing JG, Beyer EC, Saffitz JE (1998) Rapid turnover of connexin43 in 61. Elias LA, Wang DD, Kriegstein AR (2007) Gap junction adhesion is necessary for radial Circ Res – the adult rat heart. 83:629 635. migration in the neocortex. Nature 448:901–907. 47. Hervé JC, Derangeon M, Bahbouhi B, Mesnil M, Sarrouilhe D (2007) The connexin turnover, 62. Derangeon M, Spray DC, Bourmeyster N, Sarrouilhe D, Hervé JC (2009) Reciprocal an important modulating factor of the level of cell-to-cell junctional communication: influence of connexins and apical junction proteins on their expressions and Comparison with other integral membrane proteins. J Membr Biol 217:21–33. functions. Biochim Biophys Acta 1788:768–778. 48. Mathias RT, White TW, Gong X (2010) Lens gap junctions in growth, differentiation, 63. Spray DC, Duffy HS, Scemes E (1999) Gap junctions in glia. Types, roles, and plasticity. and homeostasis. Physiol Rev 90:179–206. Adv Exp Med Biol 468:339–359. 49. Pereda AE, Nairn AC, Wolszon LR, Faber DS (1994) Postsynaptic modulation of 64. Li X, et al. (2008) Connexin45-containing neuronal gap junctions in rodent retina synaptic efficacy at mixed synapses on the Mauthner cell. J Neurosci 14:3704–3712. 50. Pereda A, Triller A, Korn H, Faber DS (1992) Dopamine enhances both electrotonic also contain connexin36 in both apposing hemiplaques, forming bihomotypic gap J Neurosci – coupling and chemical excitatory postsynaptic potentials at mixed synapses. Proc Natl junctions, with scaffolding contributed by zonula occludens-1. 28:9769 Acad Sci USA 89:12088–12092. 9789. 51. Alev C, et al. (2008) The neuronal connexin36 interacts with and is phosphorylated by 65. Tuttle R, Masuko S, Nakajima Y (1987) Small vesicle bouton synapses on the distal half CaMKII in a way similar to CaMKII interaction with glutamate receptors. Proc Natl of the lateral dendrite of the goldfish Mauthner cell: Freeze-fracture and thin section Acad Sci USA 105:20964–20969. study. J Comp Neurol 265:254–274.

10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1121557109 Flores et al.